How Did Our Medical Notes Become So Useless?

In (the UK at least) your medical notes follow you as you sign on with a new GP (General Practitioner - first point of medical contact for anything non-urgent). Now knew this theoretically happened, but was quite bemused to see it happen in practice. I rescued a rat from the silken-jaws of my pseudo-homicidal cat, and the ungrateful rodent bit through half of my finger - I thought it would be sensible to have a checkup/shot. I've only got a hazy recollection of what shots I've had when, so my notes were consulted. Seemingly, according to my paper notes produced from the cabinet, 40 years ago "I was an adorable wee thing" (aged 2).

Medical notes are largely fiction IME. You should request all your medical records from any doctor you have seen and start demanding they correct them. You will be shocked at the lies they tell about you.<p>You will be even more shocked down the road when you find out how these lies can hurt you.

Medicine seems like a technological wasteland. They put &quot;notes&quot; into text boxes and call it &quot;using computers&quot;.<p>I wish I had both CS&#x2F;SE experience and Medical experience so that I could understand what keeps this field in the 1970s. I have suspicions.

There&#x27;s a recent New Yorker article by Atul Gawande covering this from a slightly different angle: <a href="https:&#x2F;&#x2F;www.newyorker.com&#x2F;magazine&#x2F;2018&#x2F;11&#x2F;12&#x2F;why-doctors-hate-their-computers" rel="nofollow">https:&#x2F;&#x2F;www.newyorker.com&#x2F;magazine&#x2F;2018&#x2F;11&#x2F;12&#x2F;why-doctors-ha...</a><p>He puts much of the blame on the generalization of computer systems, i.e. that the notes have to be readable by all sorts of staff and need to be systematized to accommodate that.

I recently built my mom a basic web app to automate building patient evaluations. She has been using paper forms with checkboxes for decades and they started requiring the info in digital format over the last few years. She had been copy and pasting from the digital version of the blank form into MS Word. It was time consuming, but still faster than typing it by hand. To be clear, a paragraph containing sentences is the format medical professionals are required to present this information in. Automating the boilerplate stuff actually gives my mom more time to type in the information that the paper forms don&#x27;t anticipate and provide actual care to patients. The problem is not automation. The problem is unscrupulous executives who maximization profits at the expense of quality of service. The examples provided in this article are unethical and anyone who is caught dumping irrelevant data into important medical records should be held responsible.

My kids&#x27; pediatrician was part of a small office, which merged with a larger office that is affiliated with a large hospital and hence uses EPIC. She used to take hand notes on paper and was quiet efficient when she was with the smaller office. After the move I have seen her struggle with EPIC, and then recently hire a medical transcription service and be followed by a person taking notes so that she can focus on the medical stuff. And, of course I have no idea if the quality of her notes has suffered.

I do NLP with medical notes. What I have observed is that the redundancy between the note and information available elsewhere in the EHR in structured formats is pretty high, as is the amount of boilerplate (e.g. section headers and list templates that may not be filled in).<p>This makes machine learning using the notes difficult since the content is so muddied up. It&#x27;s far from impossible to do useful things with them but there is a lot of noise. Still, some things require us to look at the notes. For these things we would much rather have them than not.<p>While it&#x27;s preferred to have the information entered in a structured way, doctors find that more of a hassle than entering in the data free-text and it&#x27;s also no good for retrospective analysis where we didn&#x27;t know several years ago we would be interested in something and so no structured field existed.

Here&#x27;s a fun thing about our system in the US. If you&#x27;re taken to the emergency room, they&#x27;ll ask your medical history. And not just once - they&#x27;ll ask it in the ambulance, radio it ahead to the hospital, then make you wait while they ask you the same questions again and enter them into their system <i>at</i> the hospitsl. Then the next 3 doctors or nurses who come into your room do the same thing, asking you questions you&#x27;ve already answered. Doesn&#x27;t seem to matter if it&#x27;s obvious that speaking is painful for some reason (in a case I&#x27;m describing a family member had broken ribs).

This EHR data problem is something I&#x27;ve pondered for a bit.<p>One &quot;simple&quot; solution is to have departmental standardization of note format with thoughtful inclusion of what fields are typically pertinent. This doesn&#x27;t solve the problem of care transitions but it might help standardize review in a hospital context.<p>The other thing that I&#x27;ve been pondering is something resembling a formalized data structure and language for note taking. For example, diagnosis X based on Y Z. Other probabilistic diagnosis A ruled out because not B not C yes D. Reduce free form notes to be as sparse as possible. Also there should be a reference system to point back to other notes &#x2F; lab values &#x2F; imaging which when clicked will bring up that data. Finally, a timeline which charts pertinent diagnoses, lab values, and changes over encounters. I&#x27;m not sure how viable it would be given the complexity of notes that my physician colleagues have showed me &#x2F; what I&#x27;ve seen in research, but I&#x27;m curious.

I deal with various EHRs every day of my life, and 4-digits-worth of providers. The primary barrier is that practice managers are working to get data standardized, but get pushback from the providers who are busy doing their jobs being doctors and not data scientists.<p>Some EHRs can streamline via templates and workflow. Others don&#x27;t. I&#x27;ve literally had practice admins not implement new practices that would document properly and facilitate easy reporting because they would not be able to sell the change to the providers.<p>I&#x27;m not blaming providers for the data issues. They have a job to do and they do it. It just doesn&#x27;t always get documented in a reportable manner. There needs to be an easier way to document, or some kind of Middleware that documents for them.<p>I see people a whole lot smarter than me trying to use ai to interpret notes. I personally think they are just facilitating the ongoing poor documentation problem. And things will slip through the natural language cracks, and it could be a potential health hazard.

My Primary Care Physician is unable to speak coherent sentences.<p>I&#x27;m half-joking, of course. The guy is top-rated, elected head of the state association. He&#x27;s widely-regarded as a combination of both brilliant with diagnostics and good with people. He just can&#x27;t complete a sentence when we meet.<p>Why? Because he&#x27;s got some pad he carries around that takes up all of his attention span. He comes into the room, sits down -- and there&#x27;s this struggle for his attention that I watch play out. It usually involves a lot of verbal grunts.<p>&quot;So we&#x27;ve got this .... er.. .. and it looks .. hmmmm. .... So this is....&quot;<p>This could go on for a bit. Eventually we get to either a statement or a question.<p>Frankly I&#x27;d think the guy was having some sort of mental issues if it weren&#x27;t for the facts that 1) he used to be fine before they all started carrying around pads, and 2) he&#x27;s fine outside the clinic.<p>I really hope that the tech community has helped make healthcare better. It&#x27;s certainly had an impact.

I&#x27;ve done a good bit of thinking about medical note taking.<p>To me it&#x27;s a little odd that hospitals are the ones who keep this information. I&#x27;d think that the records should belong to patients—it&#x27;s about them after all. And that the patients would provide access to doctors or hospitals.<p>Right now I&#x27;ve got a personal medical journal that has things like:<p>* Sickness - Date Range and Notes<p>* Flu Shots - Date<p>* Injuries - Date and Notes<p>* Observations - Date and Notes<p>* Blood tests - Date and Photos of Tests<p>Before my yearly checkup (or if I have to visit a clinic) I review the last entries and open them on my phone incase my doctor wants to see any of them.<p>It works well enough for me, but seems like there could be a ton of opportunity for improvement. I&#x27;d love to have a system where my doctor could be notified and comment on new notes or events. Also the ability to bring in my scale, run tracker and other fitness data.

Doc here - spot on. 25% of my day is typing, reading, deciphering, referencing. I waste so much time not providing care.

I strongly agree that EMRs are making this problem worse, not better. And doctors seem to hate quite ubiquitously the extra amount of documentation they have to do. While there is a lot of evidence in some areas that EMRs have improved quality of care and patients&#x27; lives, I don&#x27;t think this is one of them.

I worked at the VA for over 2 decades, including the transition from handwritten to electronic notes. Wow. When I retired, the nursing notes for a patient visit were usually 2-3X longer than the docs note. The nursing note was entirely populated by templated text created by a hierarchy of supervisory staff tasked with satisfying all regulatory requirements. The poor nurses would ask the patient a series of prompted questions, click on the appropriate box, and a page of templated text would emerge. Actually finding useful actionable info in the nursing notes was impossible, or close to impossible. The docs began to demand, reasonably, that actual patient problems identified by the nurse be put in a specific area of the note rather than randomly scattered throughout 4-8 pages of templated text (which without this demand could be randomly intercalated with such useful info as &quot;pt c&#x2F;o chest pain when walking, much worse&quot;, found on page 5 of 8.<p>The docs notes were, in many ways, even worse. The notes required manual typing, and many docs are not trained skilled touch typists. So the two-fingered part of the note was often very brief and succinct. The templated portions were huge - impressive reviews of systems where you could not really tell if the specific items had actually really been asked or if the template just vomited forth a page full of text for administrative review.<p>Make a doctor function as a data entry clerk and this is what happens.

The old school method of sticking the charts to the foot of the patient’s bed has the same pleasant properties that using sticky notes on a board for tasks does.<p>Anyone that’s struggled with JIRA and just said forget this due to tool impedance will have a good idea of how painful this is. In hospitals I almost never see charts pulled up except on daily rounds. Sure nurses are great and smart people and usually remember everything, but they also work hard long hours and the cost of mistakes can be very high.

My favourite citation from the link:<p>“In fact, across this same EHR, clinical notes in the United States are nearly 4 times longer on average than those in other countries”<p>It seems that the EHR isn’t the root cause of the problem.

Problem of medical notes have a name: bureaucracy. The hope is leaving it apart, and today it seems a bit utopia...<p>Also in terms of &quot;numeric&#x2F;digital&quot; vs &quot;paper&quot; the point is ignorance: how many people outside IT world (and <i>even inside</i>) do actually know enough a desktop to take <i>their own personal notes</i> in an ordered, usable and useful thing?<p>IMVHO a so small percentage that we can probably know them all by name. Just take a look at a &quot;common&quot; mailbox: most of them are an utter pile of data, few with some incoherent taxonomies, few even with the sole inbox as an archival place. Than take a look a common &quot;home directories&quot;: the very same mess. And if this is for <i>personal</i> and <i>generic</i> data do you think that those people are able to properly not only manage but share helpful information with digital systems?!

The medical journal is one of, if not the most, important inventions of modern medicine. I&#x27;m an MD and work with them every day.<p>But yes, the current infrastructure is not good.

1. Money<p>2. Legal liability (CYA)<p>3. We don&#x27;t use outcome-based medicine

I refuse to believe (beyond a bad email notification , which is not acceptable for medical information) that being elaborate or non-concise is a problem. Why is 4x more medical density a bragging point , in other words?

One &quot;solution&quot; is sometimes to use speech-to-text. The mangled results would be hilarious if they weren&#x27;t frightening.

I read number 1 —- “money” — as “doctors lie about what they are doing to get paid more by your insurance company”. Is that unfair?

<i>Spec. Off&#x27;s</i> [1]<p><pre><code> - Doctor enters your visit (encounter), cross your fingers that his system has the most up-to-date medical codes (e.g. snomed, cpt, icd, etc) - Doctors notes for that encounter are potentially entered as a text area in one system (u&#x2F;petermcneeley). Can be notes at the encounter level or for a specific diagnosis&#x2F;lab result&#x2F;etc. - The codes vs free text, is what we refer to as discreet and narrative&#x2F;free text. - The current IHE spec. [2] (last updated in 2015?), allows for codes to be interpreted from narrative text if a valid code is not provided. I think there are a few startups that have popped up here that are trying to make sense of the narrative text. - That same spec would be great if everyone followed it but they have to get their system to bend a little bit or throw an integration engine in front of the problem to play nicely with others. (u&#x2F;nradov) - Start sending those back and forth and you either end up losing those notes or butchering them up. - There are some orgs (e.g. DoD, VA, Sequoia, etc) that have everyone follow the basic requirements but then add their own flavor on top. [3] - FHIR [4] is here, but I think everyone is already scrambling or haven&#x27;t had the need to make the cut over to it until its government mandated. I have to say this has been the easiest spec. to grok, but it is still way too flexible for these companies to mess up. A cut over to FHIR for everyone should ideally be that you have to use the Hapi FHIR models [5]. Maybe have a way for Hapi to sign the models on their way out? </code></pre> Sorry, this turned out to be more of a rant, but I stare at this stuff everyday. There are a few other folks on the thread that know the industry as well that have some good info too.<p>[1] <a href="https:&#x2F;&#x2F;media1.tenor.com&#x2F;images&#x2F;af0c71048d5a130cefc335423c597ce8&#x2F;tenor.gif" rel="nofollow">https:&#x2F;&#x2F;media1.tenor.com&#x2F;images&#x2F;af0c71048d5a130cefc335423c59...</a><p>[2] <a href="http:&#x2F;&#x2F;www.hl7.org&#x2F;implement&#x2F;standards&#x2F;product_brief.cfm?product_id=258" rel="nofollow">http:&#x2F;&#x2F;www.hl7.org&#x2F;implement&#x2F;standards&#x2F;product_brief.cfm?pro...</a><p>[3] <a href="https:&#x2F;&#x2F;xkcd.com&#x2F;927&#x2F;" rel="nofollow">https:&#x2F;&#x2F;xkcd.com&#x2F;927&#x2F;</a><p>[4] <a href="https:&#x2F;&#x2F;www.hl7.org&#x2F;fhir&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.hl7.org&#x2F;fhir&#x2F;</a><p>[5] <a href="http:&#x2F;&#x2F;hapifhir.io&#x2F;" rel="nofollow">http:&#x2F;&#x2F;hapifhir.io&#x2F;</a>